Simultaneous Agricultural Failure Events and their atmospheric and climatic Drivers

Objective

Global food security is increasingly threatened by the rising frequency and geographic spread of climate extremes such as droughts and heatwaves, which often lead to simultaneous crop failures across multiple agricultural regions, or breadbaskets. These undermine food system resilience, causing significant production deficits, food shortages, price spikes, and even political instability, threatening the sustainability of social-ecological systems. Despite such severe impacts, the spatiotemporal patterns of yield failure synchronizations and their atmospheric and climatic drivers remain poorly understood, representing a critical knowledge gap for global food security governance. The SAFE-Drivers project will address this through an innovative approach integrating advanced statistical tools, machine learning, and physical modeling, with a focus on maize, wheat, soybean, and rice. The project will first employ network theory to identify synchronizations of crop failures, based on extensive yield statistics from ~20,000 political units over four decades. Then, causal discovery algorithms and Explainable Artificial Intelligence combined with Earth observations and climate reanalysis will elucidate how atmospheric precursors and local climate factors drive these synchronized failures. Finally, an atmospheric moisture and heat tracking framework will be used to explore the spatial propagation of climate extremes and their teleconnected agricultural impacts. The fellowship will also provide crucial interdisciplinary training in sustainability science, enhancing my capacity for independent research leadership in climate-agriculture interactions. Overall, the project will substantially advance scientific understanding of synchronized crop failures, enhance forecasting capabilities, and provide actionable insights for coordinated agricultural risk management, particularly benefiting food-insecure regions through improved early warning systems.